Symbolic AI (the road Dreyfus contested)

Symbolic AI was the dominant research paradigm in artificial intelligence from its founding at the 1956 Dartmouth Workshop through the collapse of the expert systems boom in the late 1980s. Its core commitment was that intelligence consists of manipulating explicit symbolic representations according to formal rules, and that building an intelligent system therefore requires specifying the right representations and the right rules. The paradigm produced significant achievements—chess programs, theorem provers, natural language parsers, expert systems—but each achievement exposed new limits, and the limits converged on the problems Dreyfus had identified in his 1965 paper: the frame problem, the common-sense knowledge problem, the embodiment problem. By the time the field transitioned to connectionism and statistical methods, the symbolic paradigm had been largely abandoned, and Dreyfus's philosophical diagnosis had been vindicated in ways his early critics had refused to anticipate.